Acquiring Diverse Skills using Curriculum Reinforcement Learning with Mixture of Experts

TL;DR

Di-SkilL is a reinforcement learning method that uses a mixture of experts and energy-based models to learn diverse, high-performing skills in complex environments, with automatic curriculum learning.

Contribution

This paper introduces Di-SkilL, a novel RL approach combining mixture of experts and energy-based models to learn diverse skills without prior environment knowledge.

Findings

Successfully learns diverse skills in robot simulations

Uses energy-based models for context distribution representation

Enables automatic curriculum learning for skill specialization

Abstract

Reinforcement learning (RL) is a powerful approach for acquiring a good-performing policy. However, learning diverse skills is challenging in RL due to the commonly used Gaussian policy parameterization. We propose \textbf{Di}verse \textbf{Skil}l \textbf{L}earning (Di-SkilL\footnote{Videos and code are available on the project webpage: \url{https://alrhub.github.io/di-skill-website/}}), an RL method for learning diverse skills using Mixture of Experts, where each expert formalizes a skill as a contextual motion primitive. Di-SkilL optimizes each expert and its associate context distribution to a maximum entropy objective that incentivizes learning diverse skills in similar contexts. The per-expert context distribution enables automatic curricula learning, allowing each expert to focus on its best-performing sub-region of the context space. To overcome hard discontinuities and multi-modalities without any prior knowledge of the environment's unknown context probability space, we leverage energy-based models to represent the per-expert context distributions and demonstrate how we can efficiently train them using the standard policy gradient objective. We show on challenging robot simulation tasks that Di-SkilL can learn diverse and performant skills.